oes ertificate of eed ffect ardiac utcomes and osts v h · 2013. 8. 6. · the james a. baker iii...

THE JAMES A. BAKER III INSTITUTE FOR PUBLIC POLICY

RICE UNIVERSITY

DOES CERTIFICATE OF NEED AFFECT

CARDIAC OUTCOMES AND COSTS?

By

VIVIAN HO, PH.D. JAMES A. BAKER III INSTITUTE CHAIR IN HEALTH ECONOMICS,

JAMES A. BAKER III INSTITUTE FOR PUBLIC POLICY, RICE UNIVERSITY;

ASSOCIATE PROFESSOR, DEPARTMENT OF ECONOMICS, RICE UNIVERSITY;

AND ASSOCIATE PROFESSOR, DEPARTMENT OF MEDICINE, BAYLOR UNIVERSITY

JULY 2007

Does Certificate of Need Affect Cardiac Outcomes and Costs?

THESE PAPERS WERE WRITTEN BY A RESEARCHER (OR RESEARCHERS) WHO PARTICIPATED IN

A BAKER INSTITUTE RESEARCH PROJECT. WHEREVER FEASIBLE, THESE PAPERS ARE REVIEWED

BY OUTSIDE EXPERTS BEFORE THEY ARE RELEASED. HOWEVER, THE RESEARCH AND VIEWS

EXPRESSED IN THESE PAPERS ARE THOSE OF THE INDIVIDUAL RESEARCHER(S), AND DO NOT

NECESSARILY REPRESENT THE VIEWS OF THE JAMES A. BAKER III INSTITUTE FOR PUBLIC

POLICY,

An earlier version of this paper has been accepted for publication and appears in the International Journal of Health Care Finance and Economics. No further reproduction of this paper is permitted without the permission of Springer. International Journal of Health Care Finance and Economics, 6:4, 300-324. ©Springer DOI 10.1007/s10754-007-9008-9 The original publication is available at www.springerlink.com. This research was supported by grant number R01 HL073825-01A1 from the National Heart, Lung and Blood Institute. I am grateful to Huifeng Yunn for helpful research assistance and to Richard Boylan, Leemore Dafny, Andrew Epstein, Ciarin Phibbs, two anonymous referees, participants at the fourth IHEA World Congress, the Department of Health Administration and Policy at the Medical University of South Carolina, and the Departments of Economics at Rice University and Texas A&M for helpful comments.


1

ABSTRACT

Several U.S. states enforce Certificate of Need (CON) regulations, which limit the number of hospitals performing open heart surgery or coronary angioplasty. CON regulations were intended to restrain cost growth and improve quality of care. This study compares mortality rates and costs for cardiac care in states with and without CON. CON appears to raise hospital procedure volume and lower the average cost of care. However, CON is associated with little reduction in inpatient mortality, and it may lead hospitals to operate on more patients than they would otherwise. The claimed welfare benefits of CON regulations require careful reconsideration. Keywords: Regulation; Certificate of Need; Outcomes Assessment; Cardiac Surgery; Costs JEL codes: I110 Analysis of Health Care Markets; I180 Health: Government Policy, Regulation, Public Health


2

I. INTRODUCTION

Several U.S. states enforce Certificate of Need (CON) regulations for cardiac care, which

limit the number of hospitals that may perform open heart surgery or coronary angioplasty.

These rules were originally implemented by regulators who argued that controlled entry into

markets for new technologies would restrain cost growth and improve the quality of care.

Limiting the number of providers would avoid unnecessary duplication of costly, highly

specialized manpower and facilities. Medical evidence suggested that minimum case loads were

essential to maintaining and strengthening the skills required to perform complex cardiac

procedures. Therefore, restricting the number of facilities that patients could choose from also

insured that each certified facility had enough patients to maintain their expertise in complex

cardiac care. Because provision of open heart surgery and angioplasty require a substantial fixed

cost investment, greater centralization of services through CON was also likely to yield

economies of scale.

Many potential providers without CON approval argue that these regulations are anti-

competitive. With fewer competitors performing open heart surgery and angioplasty in a market,

certified providers can charge higher prices for their services. Because publicly available data on

charges for cardiac procedures does not accurately reflect the prices actually paid by insurers and

patients for the care they receive, this hypothesis cannot be directly tested. However, past studies

cast doubt on the argument that CON regulations restrain cost growth or improve quality in the

cardiac care market. For instance, greater competition (as opposed to regulation of entry) has

been found to lower average expenditures per patient and mortality among Medicare patients

being treated for heart attacks in the 1990's (D. P. Kessler, M. B. McClellan 2000).

Other research suggests that the evidence supporting the clinical benefits of higher


3

procedure volume is tenuous. Analysis of longitudinal data suggests that all facilities performing

angioplasty improve over time regardless of patient volume; and within-hospital increases in

procedure volume over time provide minimal benefits in terms of patient outcomes (V. Ho

2000;Ho V. 2002). However, the impact of cardiac CON regulations on hospital procedure

volume, patient outcomes, and costs has not been studied systematically.

This study uses AHRQ HCUP data to compare patient outcomes and costs for cardiac

care in those states with and without cardiac CON regulations. The study first tests whether

hospitals in states with CON regulation perform higher numbers of open heart surgeries and

angioplasties relative to states without CON. The impact of increased procedure volume

associated with CON on inpatient mortality and average costs is then assessed. Because CON

regulations often include minimum volume standards (e.g. Providers must perform at least 200

open heart surgeries per year), these rules may lead providers to perform surgery on patients that

they would otherwise have treated with medications. Therefore, the analysis also tests whether

states with CON have a greater number of individuals overall receiving these cardiac

interventions.

The results suggest that CON regulations increase patient volume. However, CON

regulations are associated with only small reductions in inpatient mortality for open heart surgery

patients, and they have no tangible effect on inpatient mortality for patients undergoing

angioplasty. The regulations lead to noticeable reductions in average cost. However, CON also

increases the propensity to perform cardiac surgery, which may raise total expenditures for these

procedures.

The results have important implications for understanding the role that regulations can

play in controlling health care costs and influencing the quality of patient care. The CON


4

regulations for cardiac care appear to have only a marginal impact on patient mortality. The

regulations have been more effective in reducing average costs through economies of scale.

However, increases in the propensity to perform cardiac procedures with CON may still lead

to higher expenditures for the health care system as a whole.

Section 2 provides background on cardiac CON regulations and previous related

research. Section 3 describes the dataset to be used for analysis and provides descriptive

statistics. Section 4 outlines the models to be estimated, and section 5 presents results.

Conclusions are provided in Section 6.

II. BACKGROUND

The National Health Planning and Resources Development Act of 1975 (Public Law

93641) authorized funding for state CON programs. CON programs required that certain

proposed capital expenditures, changes in health facilities or services provided, or purchases of

major medical equipment were subject to the review and approval of a designated state agency.

If a state did not implement a CON program, then no providers in the state were eligible to

receive funds under the Public Health Service Act for the development, expansion, or support

of health resources in the state.

CON programs were originally adopted in part to achieve cost containment. However, a

study of hospital regulation over the period 1966 to 1982 found that CON appears to increase per

capita hospital expenditures in the long run (J. A. Lanning et al. 1991). Using annual state-level

data collected from the American Hospital Association and other sources, Lanning et al. found

that per capita hospital expenditures were 20.6 percent higher in states with CON versus states

without CON. The authors suggest that CON programs may act to protect inefficient hospitals


5

from competition.

More recent studies have been able to analyze health care expenditures after the repeal or

sunset of various state CON laws. Conover and Sloan examined annual state-level data from

1980 to 1993 to test whether lifting of CON regulations led to changes in per capita hospital

spending (C. J. Conover, F. A. Sloan 1998). This study found no evidence of lower hospital costs

per capita in states with mature CON programs; or a surge in hospital costs following removal of

CON regulations. Similarly, Grabowski et al. found no evidence of increased nursing home

expenditures in states which repealed CON regulations or construction moratoria (D. C

Grabowski et al. 2003).

These studies provide no support for the argument that CON laws are helpful in

restraining expenditure growth. CON laws may in fact increase market power, enabling high

volume providers to raise the prices they charge for procedures. CON laws have been found to

deter entry and allowed hospitals to raise prices 4.0 to 4.9 percent (M. Noether 1988). However,

none of these studies have examined the impact of CON regulations on expenditures for a

particular treatment class, such as cardiac surgery. It is particularly interesting to study the

impact of CON for cardiac surgery, because these patients represent an extremely high

proportion of total inpatient care for large hospitals; and caring for them can be extremely

profitable. In addition, with the exception of Grabowski et al., these studies do not analyze the

impact of CON laws on both the total quantity of services provided and average costs per patient.

Although CON may be successful in helping unit costs to decline (through economies of scale),

an offsetting increase in the total number of procedures performed may be cause for concern, if

the marginal benefit of these procedures is low. Although this study does not directly address the

issue of the marginal benefit of increased access to cardiac surgery, it does attempt to measure


6

the potential for changes in the total population receiving intervention as a result of CON

legislation.

In 1978, the Federal government issued the National Guidelines for Health Planning,

which included standards on the minimum numbers of open heart procedures and cardiac

catheterizations that should be performed annually in cardiac care facilities. The guidelines

stated that there should be a minimum of 200 open heart procedures performed annually, within

three years after initiation, in any institution in which open heart surgery is performed in adults.

This recommendation was made in order "to maintain quality of patient care and make most

efficient use of resources." In addition, the guidelines stated, "In order to prevent duplication of

costly resources which are not fully utilized, the opening of new units should be contingent upon

existing units operating, and continuing to operate, at a level of at least 350 procedures per year."

Open-heart surgery is performed while the bloodstream is diverted through a heart-lung

bypass machine. The overwhelming majority of these operations are coronary artery bypass graft

(CABG) procedures. Standards were set for cardiac catheterization as well: "There should be a

minimum of 300 cardiac catheterizations, of which at least 200 should be intracardiac or

coronary artery catheterizations, performed annually in any adult cardiac catheterization unit

within three years after initiation." This standard effectively regulates percutaneous transluminal

coronary angioplasty (PTCA), because a cardiac catheterization laboratory is required to perform

angioplasty. State CON review criteria were expected to reflect these national standards.

The Federal law supporting CON expired in 1986, leading many states to discontinue

their CON programs. Currently, 35 states and the District of Columbia fund and administer a

CON program. Of these, 28 continue to explicitly regulate cardiovascular services. A 2000

Maryland Health Care Commission survey found that no state reported studying the effect of


7

its repeal (Maryland Health Care Commission 2000). Even for those CON states without

explicit requirements for cardiac services, the general presence of CON regulations (e.g.

restrictions on bed supply) may limit the number of hospitals in the market, which could lead

to higher volumes for CABG and PTCA among existing providers. However, this hypothesis

has not yet been examined in the literature.

Meanwhile, applications to establish new cardiac programs in states which maintain CON

regulations are attracting growing attention in the press (B. Wysocki 2002;S. Hensley 1997;K.

Pallarito 1998). Increased competition in the health care market along with the high profitability

of performing cardiac surgery imply great financial returns to those facilities which are able to

gain certification or prevent the entry of new competitors in a local market. For example,

Coliseum Health System’s application to perform high-end cardiac care in Macon, Georgia has

been denied, granted or appealed seven times since 1999, in part due to legal actions by Macon’s

existing provider of these services, Medical Center of Central Georgia (B. Wysocki 2002).

A recent study published examined the short-term impact of the termination of cardiac

surgery CON laws in Pennsylvania in 1996 (J. L. Robinson et al. 2001). In the three years prior

to the termination of CON in Pennsylvania (1994-1996), the number of open-heart surgery

programs rose from 43 to 44. In contrast, in the three years following CON termination, the

number of programs increased from 44 to 55. The study found no significant change in the

inpatient mortality rate for CABG. The impact of CON termination on mortality for PTCA was

not reported. The authors speculate that the sharing of surgeons between open-heart surgery

programs established before and after CON termination may have helped to prevent an increase

in inpatient mortality with the opening of 11 new cardiac programs.

In contrast, a more in-depth analysis of Medicare data found a substantial association


8

between cardiac CON laws and reduced mortality for patients undergoing CABG. Vaughan-

Sarrazin et. al. used Medicare Provider and Analysis Review (MedPAR) Part A data from 1994

through 1999 from all 50 states to examine the impact of CON laws on delivery of care for

CABG patients (M. S. Vaughan-Sarrazin et al. 2002). The analysis found that hospital volumes

in the 26 states with continuous CON was 84 percent higher than in the 18 states with no CON

(191 cases per year vs. 104 cases per year). The study also reports the remarkable finding that

risk-adjusted in-hospital mortality was 22 percent higher for patients in states with no CON

regulation versus states which maintained CON.

The authors acknowledge an important limitation in their analysis; any associations

between CON regulation and study outcomes may represent confounding due to other factors

which may differ according to CON status. That is, the 22 percent mortality differential between

CON and no-CON states may in fact be due to other factors which are correlated with the extent

of regulation in a state (e.g. level of managed care penetration, local physician practice styles,

etc.). In contrast, this study will employ fixed effects in all regression analyses to control for

unobserved hospital and state characteristics. Although the Vaughan-Sarrazin et al. analysis is

based on data from all 50 states, the Medicare dataset lacks information on patients under age 65,

as well as those enrolled in Medicare managed care. Therefore, the differences in hospital

volume and mortality identified in this study may be upward biased due the higher rate of

managed care penetration in states without CON regulation in their sample (V. Ho 2003).

Although this study lacks data from all states, the sample includes patients of all ages, regardless

of insurance type. It is important to test whether these notable differences in volume and

outcomes for CABG persist with these differences in both the sample and method of analysis.


9

III. DATA AND DESCRIPTIVE STATISTICS

This study uses data from the AHRQ HCUP Nationwide Inpatient Sample (NIS) database

to compare hospital procedure volumes and costs for PTCA and CABG in states with and

without cardiac CON laws. The NIS data is analyzed for the years 1988 to 2000. Ideally, one

would conduct an analysis of changes in procedure volume and costs in states before and after

repeal of CON legislation, using states that never repealed CON as a control group.

Unfortunately, all of the states in the HCUP sample except one discontinued their cardiac CON

program prior to 1988. Therefore, analyses in this study will focus on examining differences in

hospital procedure volume or costs, as a function of the number of years since CON legislation

was repealed in a given state.

The HCUP NIS data is a stratified random sample of community hospitals in the United

States. The dataset contains a 20 percent sample of acute care hospitals in each of five strata

(geographic region, ownership, location, teaching status, and bedsize). The first release of the

NIS for the years 1988 to 1992 contained information from 11 U.S. states, and the 2000 release

contains information on over 1,000 hospitals in 28 states. Although the sampling strategy for the

NIS re-sampled many hospitals over several years, very few facilities are included for all 13

years. On average there are four years of data for each hospital in this study.

The NIS contains patient-level clinical and resource use information included in a typical

discharge abstract. All discharge abstracts from each sample hospital are included, so that one

can obtain accurate counts of PTCA and CABG volume for each facility. All patients with a

procedure code for CABG (ICD-9-CM 36.1x) or PTCA (ICD-9-CM 36.01, 36.02, or 36.05) were

included in the sample for this study.

Information on CON status for each state was drawn from the 2000 Maryland Health


10

Care Commission survey. The survey lists the 28 states which maintained a cardiac CON

program through the year 2000 (18 are in the NIS sample), as well as the eight states which

never implemented a CON program (2 are in the NIS sample). The survey also lists the 15 states

which repealed their cardiac CON legislation and the year in which the repeal occurred (seven

are in the NIS sample). See the Appendix A for a summary of CON programs as reported by the

Maryland survey.

Figure 1 graphs mean PTCA and CABG volume by year for hospitals in states with and

without cardiac CON. Note that in both cases that procedure volume in CON and non-CON

states were remarkably similar in the years just following repeal of the Federal CON

legislation. However, procedure volume begins to diverge between CON and non-CON states

in 1992, and mean hospital volumes in CON states remain higher through 2000. These data are

consistent with the hypothesis that CON restricts entry, allowing CON hospitals to grow larger

on average. One may be concerned that this observed difference is attributable to the increasing

number of states and hospitals added to the NIS sample over time. The graph was repeated

including only hospitals in states that were in the NIS for all years from 1988 to 1992. The

pattern in the graph remains almost the same.1

Figure 1 excludes data from Pennsylvania, which repealed its cardiac CON legislation in

1996. Procedure volumes by year are reported for Pennsylvania in Figure 2. Note the dramatic

decline in average procedure volume for both PTCA and CABG after 1996.2 These changes in

average volume are consistent with a previous study which noted that the average number of

open-heart surgery programs in Pennsylvania increased from 44 to 55 between 1997 and 1999 (J.

1 Results are available from the author upon request. 2 An unusually large rise in PTCA and CABG volume occurs just prior to CON repeal in Pennsylvania in

1996, as well as a relatively large drop in average costs for CABG. Hospitals with CON may have aimed to boost


11

L. Robinson et al. 2001). Figure 3 presents graphs of the average costs per patient in hospitals

performing PTCA and CABG. Costs for each patient were derived by multiplying the reported

charge for each patient by a hospital-specific cost-to-charge ratio obtained from Medicare Cost

Reports.3 These costs were then deflated by the All-Urban Consumers CPI to reflect 2000

dollars. Past studies of disease-specific costs have also relied on aggregate hospital cost-to-

charge ratios rather than department-level data (D. Meltzer et al. 2002;D. M. Cutler, R. S.

Huckman 2003). As one of these studies notes, analysis of a specific treatment with large fixed

costs such as CABG and PTCA yields a more homogeneous sample of hospitals, which is likely

to mitigate any bias created by the use of hospital-level data. Further, costs computed using

aggregate hospital cost-to-charge ratios have been found to be generally within 10 percent of

costs obtained from department-level cost data (D. Dranove 1995).

Note that average costs tend to be lower in hospitals in states with CON, particularly for

CABG procedures. If there are economies of scale in performing PTCA and CABG, then higher

procedure volumes in CON states may lead to lower costs of care per patient.4 Figure 4 presents

average costs for hospitals in Pennsylvania. The graph for PTCA suggests a sharp increase in

PTCA costs after the repeal of CON legislation, although the effect is not lasting. Average costs

for CABG also rise somewhat after the repeal of CON legislation; although again the effect does

not appear to persist.

their cardiac procedure numbers in anticipation of repeal, in an attempt to deter potential entrants. Volume growth has been identified as a method of strategic entry deterrence in previous research (L. Dafny 2005).

3 The hospital cost-to-charge ratio was calculated excluding outpatient service cost centers, as well as inpatient service cost centers that were not likely to be used by cardiac care patients (e.g. nursery, delivery room, and labor room costs).

4 Average costs for both PTCA and CABG declined over the period 1988 to 1996 for both CON and non-CON states. Average hospital volume for both of these procedures tended to increase for all hospitals during this time period. Because annual and cumulative hospital procedure volume are highly correlated, prior research has not been able to determine whether this decline is attributable to economies of scale or learning by doing (Ho V. 2002).


12

IV. MODEL SPECIFICATION

The previous figures do not control for differences in patient populations across states

which may explain differences in procedure volume for PTCA and CABG. The following

regression specification is estimated to identify the effect of CON legislation and other factors

on hospital PTCA procedure volume:

(1) log(PTCA)ht = f (yrsnoCONht, Zht, Yeart, θh)

where log(PTCA)ht is the natural log of the number of the of PTCA procedures performed by

hospital h in year t. The explanatory variable yrsnoCONht is the number of years since state CON

legislation has been repealed for hospital h in year t. This variable is set equal to 0 for hospitals

which maintained cardiac CON legislation through 2000. For Oregon and Wisconsin, which

never imposed cardiac CON legislation, the year of repeal was coded as 1987, when the general

Federal CON program was dismantled.5

A vector of control variables Zht are hypothesized to capture differences in demand for

cardiac procedures across hospitals. These include the natural log of the population age 65 plus

per square mile and the HMO penetration rate in hospital h’s county of operation in year t.6

State-level annual data on smoking rates, per capita income, and the percent of the population

uninsured were also included in preliminary regressions. However, per capita income and

uninsurance rates are excluded from the reported estimates, because they had poor explanatory

power and did not influence the other regression estimates. See Appendix B for a list of

summary statistics for these data and data sources.

5 In the absence of specific CON rules for cardiac care, general Federal CON legislation in these two states

was likely to restrict the number of providers performing PTCA and CABG up through 1987. All of the estimation results were robust to exclusion of these two states. In all but one case, the coefficient estimates remained unchanged up to at least the second decimal point, and in most cases up to the third decimal point. Precision changed only slightly.


13

Hospital-level counts of the number of heart attack patients and the number of patients

admitted with chest pain were also available. However, preliminary regressions indicated

substantial multicollinearity between these two variables and the population per square mile.

Therefore, only the population per square mile is included in the regressions.

The regression includes a vector of year dummies and a vector of hospital-specific fixed

effects. Thus, the coefficient on yrsnoCON can be interpreted as the within-hospital increase in

PTCA volume associated with each additional year past repeal of state CON legislation. A

similar regression is estimated to determine the association between the number of years since

CON repeal and hospital CABG volume. As explained previously, only Pennsylvania switched

its cardiac CON status during the sample period. Therefore, the sample size is insufficient to

estimate hospital-specific fixed effects regressions with only a dummy variable for CON as the

explanatory variable of interest.

Specifying a linear relationship between the number of years since CON repeal and

hospital volume procedure volume seems reasonable, given that Figure 1 suggests a gradual

divergence in both PTCA and CABG volume between CON and non-CON states in the 1990’s.

The figures are consistent with the hypothesis that limited entry in CON states enabled hospitals

to markedly increase procedure volume year after year, while volume in non-CON states grew

more slowly or remained constant.7

I also test whether changes in hospital procedure volume attributable to CON

legislation have a noticeable impact on inpatient mortality for PTCA and CABG. Given that

6 County-level information was drawn from the Area Resource File. Six states did not report hospital

identifiers in the NIS (GA, HI, KS, SC, TN, and TX). State-level data was assigned in place of county-level information for hospitals in these states.


14

higher procedure volume is associated with better patient outcomes, it is hypothesized that

an increase in hospital procedure volume associated with CON will be associated with lower

inpatient mortality. To test this hypothesis, I estimate regressions of the following form:

(2) (RAMRPTCA)ht = f ((PTCA Volume)ht , Yeart , θh )

where (RAMRPTCA)ht is the risk-adjusted inpatient mortality rate for patients undergoing PTCA

in hospital h in year t, and (PTCA Volume)ht represents the number of PTCA procedures

performed by hospital h in year t. The remaining variables on the right-hand side of this equation

have definitions identical to those provided for equation (1).

The data are aggregated to the hospital-year level for estimation of the mortality

regressions in order to obtain results which are interpretable for policy purposes. Estimation of

the relationship between hospital volume and mortality at the patient level would require a logit

specification. However, one cannot compute partial effects from a hospital fixed-effects logit

equation, because the distribution of the unobserved hospital heterogeneity component is

unknown (J. M. Wooldridge 2002). Aggregation to the hospital level yields the mortality rate as

the dependent variable, so that fixed effects estimates are readily interpretable.

An RAMR is calculated as the hospital’s observed mortality rate for a given procedure in

a given year, divided by its expected mortality rate, all multiplied by the entire sample’s average

mortality rate in that year. This figure provides an estimate of what the hospital’s inpatient

mortality rate would have been if it had served a mix of patients identical to the entire sample’s

mix.

The hospital’s expected inpatient mortality in a given year is calculated using the

coefficient estimates of a logistic regression of inpatient mortality on patient-specific

7 The data for Pennsylvania in Figure 2 also appears consistent with this hypothesis. After repeal of CON


15

characteristics. Thus, a separate logistic regression is estimated for each year in the sample,

where the unit of observation is a patient in that year and the dependent variable equals 0 if the

patient was discharged alive and equals 1 if the patient died in hospital. Separate logistic

regressions are also estimated for PTCA and CABG patients. The patient characteristics included

in the logistic regression are: age (dummy variables for ages 65 to 69, 70 to 74, 75 to 79, 80 to

85, and 85+ versus those under age 65), as well as indicator variables for female, black,

emergency admission, transfer from another hospital, primary diagnosis of acute myocardial

infarction (AMI, commonly known as a heart attack). Indicator variables are also included for

the individual comorbidities that comprise the Charlson comorbidity index (P. S. Romano et al.

1993), which are: prior AMI, peripheral vascular disease, dementia, chronic obstructive

pulmonary disease, rheumatologic disease, liver disease (mild), liver disease (moderate/severe),

diabetes (mild/moderate), diabetes with complications, kidney disease, cancer, and metastatic

solid tumor. For the regressions for PTCA patients, indicator variables for stent insertion and

performance of multiple-vessel PTCA were also included.

As with prior studies, I hypothesize that the coefficient on procedure volume will be

negative. That is, higher procedure volume will be associated with lower inpatient mortality.

Specification tests are run to determine whether the relationship between hospital volume and

risk-adjusted mortality is best represented with the log of volume or polynomials of hospital

volume.8 The specification with the best fit is reported in the tables.

Regressions with average hospital costs per patient for both PTCA and CABG are also

in 1996, average procedure volume for both PTCA and CABG fell dramatically, then remained relatively constant, similar to non-CON states in Figure 1.

8 The standard errors for volume in logs and polynomials are compared in regressions including both volume specifications. In addition, predicted mortality based on log volume is added as a regressor in a mortality regression with polynomials of procedure volume to determine whether it yields any additional explanatory power. Likewise, predicted mortality based on polynomials of volume is added as a regressor in a mortality regression with volume expressed in logs to evaluate its explanatory power.


16

estimated. The same patient characteristics which are used to construct risk-adjusted mortality

rates are included as explanatory variables in the cost regressions for PTCA and CABG patients.

Patient-specific factors such as age, gender, presence of comorbidities, and urgency of the

procedure are all likely to affect costs.

The presence of minimum procedure volume standards in cardiac CON legislation may

lead hospitals to perform more cardiac procedures than they otherwise would have. To test this

hypothesis, one needs accurate measures of statewide procedure volume for both PTCA and

CABG. One could then determine whether states that maintain CON rules experienced higher

growth in procedure volume versus those states that repealed the rules. Unfortunately, the HCUP

NIS does not contain information on all hospitals in any state, so that this dataset cannot be used

to obtain accurate statewide procedure volume estimates. However, the AHRQ maintains an

interactive tool on its website which allows one to determine the total number of PTCA and

CABG procedures performed by all hospitals in select states that participated in the NIS for the

years 1997 to 2002. I supplemented this information with PTCA and CABG counts based on

comprehensive hospital discharge abstract covering the years 1988 to 1996 for California,

Florida, New Jersey, and New York, as well as data from 1990 to 2001 for Colorado and

Pennsylvania.9 The resulting sample contains 145 state-year observations.

Regressions are then estimated to identify the impact of number of years beyond repeal of

CON legislation on procedure volume at the state level. These regressions include year dummies,

the log of population age 65 plus per square mile, the HMO penetration rate, the smoking rate,

9 Obtaining comprehensive discharge abstract data from individual states is time consuming and costly. I

chose to collect data from four large states, which represent 27 percent of the population in the United States. Because so few states in the NIS discontinued their cardiac CON programs, I also purchased the state-level data which was available for this time period from Colorado and Pennsylvania.


17

per capita income, and the percent uninsured at the state level. State-level estimates of the

number of heart attack patients and the number of patients admitted with chest pain each year

were also available. However, similar to the hospital-level analyses, preliminary regressions

indicated substantial multicollinearity between these two variables and the population per square

mile. Therefore, only the population per square mile is included in the regressions.

The state-level results are first reported including state fixed effects, then with first-

differencing. Correlation of the error term with the explanatory variables in any time period

(current, past, or future) leads to inconsistent estimates (J. M. Wooldridge 2002). Therefore,

for each specification, a Wald test was performed to test for strict exogeneity of the

explanatory variables with respect to the error term.

The fixed effect estimates of the determinants of annual procedure volume and costs are

estimated using the areg command in Stata 8.0. This approach allows for the estimation of fixed

effects, as well as heteroskedasticity-robust standard errors which also account for the correlation

of the error terms across time within hospitals. Regressions estimated at the hospital level

incorporate hospital-level fixed effects; and regressions at the state level incorporate state-level

fixed effects.

In cases where the dependent variable is risk-adjusted mortality, the regressions are first

estimated by OLS with standard errors robust to clustering within hospitals, but no hospital-

specific fixed effects. These estimates provide a standard of comparison to volume-outcome

effects for CABG and PTCA which are commonly reported in the medical literature. The

effects of procedure volume on mortality with hospital-specific random effects, then fixed

effects are also reported. Although the risk-adjusted mortality rates control for observable


18

differences in casemix across hospitals, the discharge data may not fully capture more detailed

clinical characteristics of patients that may be correlated with hospital volume and risk of

death. Regression estimates based on fixed effects will provide consistent estimates of the

effect of procedure volume on mortality, whether or not unobservable differences in patient

casemix are correlated with volume. However, random effects estimates will yield smaller

standard errors if unobserved heterogeneity across hospitals is uncorrelated with hospital

volume. Therefore, Hausman tests are performed to determine whether the random effects or

fixed effects specification is more appropriate when examining the impact of hospital volume

on mortality for both PTCA and CABG.

Past studies suggest that hospital procedure volume may be an endogenous regressor. The

association between high procedure volume and favorable patient outcomes may be driven in

part by selective referral. For complex procedures, doctors may refer their patients to facilities

that they perceive to provide higher quality care. If hospital volume is endogenous in regressions

analyzing the determinants of inpatient mortality, then an alternative estimation strategy is to

estimate a two-stage instrumental variables equation. The number of years since CON repeal is a

likely candidate for an instrument that predicts procedure volume, but is unlikely correlated with

unobserved differences in mortality across hospitals. The population per square mile in a

hospital’s county and the number of patients admitted with a heart attack are also hypothesized

to influence the demand for cardiac care and therefore serve as instruments. If in fact hospital

procedure volume is not an endogenous regressor in the mortality regressions, then this approach

will yield inefficient estimates. Hausman specification tests were conducted to test for the

endogeneity of procedure volume. In each case, one could not reject the null hypothesis that the

preferred random or fixed effects estimates treating volume as exogenous were consistent and


19

efficient. Therefore, only these estimates are reported in this study.

V. RESULTS

Table 1 contains regression estimates for the determinants of hospital PTCA and CABG

procedure volume. In general, later years are associated with increases in procedure volume for

both PTCA and CABG. County population age 65 plus per square mile increases volume for

these procedures, although the coefficients on other factors hypothesized to influence demand

(HMO penetration and the smoking rate) are imprecisely estimated.10

The coefficient on yrsnoCON is negative and precisely estimated in both the PTCA and

CABG regressions, although the coefficient on yrsnoCON has a p-value of only 0.07 in

explaining hospital CABG volume. The average value of yrsnoCON for the states in the sample

which repealed CON legislation is equal to 14 years in 2000. The coefficient estimate on

yrsnoCON in the PTCA regression is equal to -0.033. This estimate implies that on average,

PTCA volume for hospitals in states which repealed cardiac CON legislation was 37 percent

lower than it would otherwise have been by the year 2000 if CON rules had been maintained

(using the formula ) (J. M. Wooldridge 2000). Similarly,

the estimates for the CABG regression suggest that average hospital CABG volumes were 27

percent lower by 2000 than they otherwise would have been.

Table 2 contains regression estimates of the relation between procedure volume and

hospital risk-adjusted mortality for both PTCA and CABG procedures. Due to space constraints,

the first-stage patient-level estimates of inpatient mortality regressed on patient characteristics

10 Nevertheless, HMO penetration and the smoking rate are included in the regression, because of the

strong priors that these variables should influence demand. The coefficients on other explanatory variables are virtually the same in specifications without these variables.


20

are excluded from the tables.11 The estimates are similar to those reported in previously

published studies. Older patients have higher mortality rates. Black patients, urgent/emergency

admissions, and those with severe comorbidities are also more likely to die in hospital.

The first three columns of Table 2 report the association between the annual number of

PTCA procedures performed by a hospital and its risk-adjusted mortality. The OLS and random

effects estimates suggest that higher hospital procedure volume is associated with lower

mortality rates for PTCA patients. However, the coefficient on volume is imprecisely estimated

when hospital-specific fixed effects are included. A Hausman test rejects the null hypothesis that

the random effects estimates are consistent. This test result suggests that unobserved differences

in patient casemix across hospitals are correlated with both procedure volume and mortality.

This might be the case if, for example, hospitals that perform higher numbers of PTCA

procedures also attract greater numbers of patients with unstable angina or lower ejection

fractions. These factors have been demonstrated in previous clinical studies to be risk factors for

increased inpatient mortality among PTCA patients (S. E. Kimmel et al. 1995;E. L. Hannan et al.

1997). Therefore, although CON appears to lead to a sizeable increase in average hospital PTCA

volume, there is no evidence that beneficial reduction in inpatient mortality accompanies this

volume change.

The last three columns of Table 2 present regression estimates of the impact of

hospital procedure volume on risk-adjusted mortality rates for CABG patients. Again, the

regression estimates including hospital-specific effects yield an imprecise estimate for the

log(volume) coefficient. However, in this case, a Hausman test indicates that one cannot

reject the null hypothesis that the random effects estimates are consistent and efficient

11 The results are available from the author upon request.


21

(χ2(13)=16.1, p=0.24). The volume coefficient in the random effects risk-adjusted mortality

regression for CABG is equal to -0.004, indicating a negative and precisely estimated

association between increased CABG procedure volume and inpatient mortality. It is

notable that the magnitude of the volume coefficient in the random effects regression is

two-thirds the size of the volume effect obtained from estimating pooled OLS. Therefore,

previous clinical studies based on OLS estimates may have overestimated the benefits of

increased volume in reducing inpatient mortality.

To examine the magnitude of this effect in the context of CON legislation, one can

substitute equation (1) into equation (2). The RAMR associated with an increase in the number

of years past repeal of CON then becomes the coefficient on yrsnoCON in equation (1)

multiplied by the coefficient on log(CABG Volume)ht in equation (2), multiplied by yrsnoCON.

One finds that for an average hospital in a state 14 years beyond repeal of CON rules, inpatient

mortality rates for CABG are 0.1 percentage points higher than they otherwise would have been.

In the year 2000, 29,194 patients in the NIS sample underwent CABG in states which had

repealed CON rules. In these states, the average risk-adjusted mortality rate was 3.9 percent in

2000. These estimates suggest that retaining CON would have led to 29 fewer inpatient deaths

overall in that year.

One can apply a similar exercise to the PTCA results to assess whether the imprecise

coefficient estimates relate to the absence of a tangible effect of volume on mortality, or instead

an insufficient sample size. The lower bound of the 95 percent confidence interval for PTCA

volume in the fixed effects regression in Table 2 is -0.0038. Let this value represent the largest

potential effect that hospital volume could reasonably have in reducing mortality for PTCA

patients. One then infers that 14 years after CON repeal, the mortality rate for PTCA patients


22

would be .2 percentage points lower if CON had instead been retained. In 2000 there were

51,539 PTCA patients in non-CON states in the NIS sample, and their risk-adjusted mortality

rate was 1.6 percent. Therefore, even at the 95 percent confidence limit, the estimates suggest

that retaining CON would lead to 103 fewer deaths, a relatively small amount.

Table 3 contains regression estimates of the relation between repeal of CON legislation

and mean hospital costs. The estimates indicate that increasing PTCA volume is associated with

lower average costs per patient. This result is consistent with the hypothesis of substantial

economies of scale in the performance of these procedures. Specification tests indicated that

polynomials of hospital volume to the third power (rather than log volume) provided the best fit

in explaining the cost of performing CABG. The estimates indicate that within-hospital increases

in CABG volume are associated with declining average costs up to an increase of 852

procedures. The mean one-year increase in hospital CABG volume is 15 procedures, and the

largest one-year increase is 490 surgeries. In addition, the largest within-hospital increase in

CABG volume over the entire time period is 834 procedures. Therefore, the estimates are

consistent with economies of scale in performing CABG over the range of procedure volumes

observed in the sample.

Applying the same exercise described above, the volume estimate in the PTCA cost

regression suggests that on average, hospitals in states that repealed cardiac CON legislation had

mean patient costs in 2000 which were 3.0 percent higher than they otherwise would have been.

Among states in the sample without CON legislation in 2000, mean CABG volume equalled 300

procedures. The results in Table 1 suggest that average procedure volume for CABG is 27

percent lower than it would have been in 2000 if these states had maintained CON rules;

implying that average CON volume would have equaled 411 procedures per year with CON. The


23

cost estimates in Table 3 suggest that a within-hospital increase in CABG volume of 111

procedures (from 300 to 411) is associated with a 6.7 percent decrease in average costs. Given

the large number of patients undergoing these procedures, these estimates suggest that by

lowering average hospital procedure volume, repeal of CON legislation potentially has a

noticeable impact on overall expenditures.

Regressions were also estimated with state-level volumes for PTCA and CABG as the

dependent variables. The results are reported in Table 4. For the fixed effect estimates, a test of

strict exogeneity for the number of years since CON repeal and the population per square mile

involves performing a Wald test on one-year leads of these variables when added to the reported

specifications. For the first-difference regressions, a Wald test of the significance of number of

years since CON repeal and the population per square mile in levels when added to the reported

specification is used to test for strict exogeneity.

For the state PTCA volume regressions, one cannot reject the hypothesis of strict

exogeneity of the explanatory variables (F(6,20)=0.81), and therefore the fixed effects estimates

are the most efficient. Each year beyond CON repeal appears to lower the number of PTCA

procedures performed in the state by 3.8 percent. For the CABG regressions, one rejects the null

hypothesis of strict exogeneity in the fixed effects estimates. However, one does not reject the

hypothesis of strict exogeneity of the explanatory variables in the first-difference regression.

The results suggest that each year beyond CON repeal reduces statewide CABG surgeries by 1.6

percent.

For both PTCA and CABG, the repeal of CON appears to be associated with a

decrease in the total number of procedures performed statewide. The estimates suggest that 14

years after repeal of cardiac CON laws, statewide procedure volumes for PTCA were 41 percent


24

lower than they otherwise would have been; and statewide procedure volumes for CABG were

18 percent lower.

The sizeable difference in statewide PTCA volume for states which repealed CON laws

may be attributable to the limited sample. California and Colorado are the only states which

repealed CON and are represented in the sample prior to 1997.12 Between 1989 and 2001,

California’s PTCA volume increased from 22,924 to 49,613 procedures per year, or 116 percent.

Colorado’s PTCA volume rose from 1,996 to 6,046 procedures per year between 1990 and 2001,

or 203 percent. In contrast, Florida, New Jersey and New York are represented in all of these

sample years and maintained CON regulations. In these states, the average PTCA procedure

volume rose from 8,353 to 34,522 procedures per year between 1989 and 2001, or an increase of

313 percent. Analysis of data from a wider range of states would be helpful in confirming this

finding.

However, it is also interesting to note that between 1996 (when Pennsylvania eliminated

it’s CON rules) and 2001, total PTCA procedure volume in Pennsylvania rose 23 percent.

During this time period, statewide PTCA volume rose an even higher average of 57 percent for

Florida, New Jersey, and New York. Between 1990 (the first year for which Pennsylvania data is

available) and 1995, rates of growth for PTCA were more similar for Pennsylvania versus these

3 states that retained CON; 74 percent versus 90 percent, respectively. These data are also

consistent with the hypothesis that CON rules lead to higher statewide performance of PTCA.

The reported specifications of the association between years since CON repeal and

procedure volume at the hospital and state level are linear, suggesting that repeal of CON has a

persistent effect on procedure growth rate through time. To test this assumption, the number of

years since CON repeal for all states that repealed CON was divided into tertiles, then quartiles.


25

The tertiles/quartiles were used to estimate splines to determine whether the relationship between

years since CON repeal and PTCA/CABG volume weakened for the years that were furthest

from the repeal (i.e. closest to the present).

I tested the hypothesis that the slope of the relationship between volume and years

since CON repeal was constant by testing whether the coefficients on the two highest tertiles

or three highest quartiles were jointly equal to 0. The results are reported in Appendix C. In

seven of the eight cases, one cannot reject the hypothesis that the relationship between years

since CON repeal and procedure volume is constant. For state-level PTCA volume, however,

the results suggest that the effect of CON repeal weakens at the third quartile, or 12 years

after repeal. CON repeal still is associated with lower statewide PTCA volume after 12 years,

although the effect is only about one quarter of its magnitude in the first six years.

I also estimated the hospital-level and state-level PTCA and CABG volume regressions

adding a quadratic of number of years since CON repeal, then a quadratic and cubed term. In

each case the higher order terms were imprecisely estimated, and the standard error for the

coefficient on the linear term increased in magnitude. For example, for the quadratic of number

of years since CON repeal, the p-values ranged from .189 to .790.

Finally, I re-estimated the hospital and state-level volume regressions using data only

through 1995, to examine whether a relatively long follow-up period could be introducing

unrelated time series variation. The coefficient on the number of years since CON repeal was -

.026 (p=.21) and -.014 (p=.003) in the hospital-level PTCA and CABG regressions, and -.037

(p=.007) and -.030 (p=.08) in the state-level PTCA and CABG regressions. One would expect

some drop in precision due to the smaller sample size. However, in each case the direction of the

effect remains unchanged, and the precision of the estimates remains relatively high.

12 The AHRQ HCUP website only reports statewide data by procedure volume from 1997 onwards.


26

VI. CONCLUSIONS

The reported estimates suggest that substantial declines in average hospital PTCA and

CABG procedure volume have resulted in states which repealed cardiac CON legislation.

Although not explicitly tested in this paper, these declines are likely due to the larger number of

new providers entering those states which do not impose minimum volume standards for these

procedures through CON laws. Reductions in average hospital volume associated with the

absence of CON have a detrimental impact on mortality rates for CABG patients, although the

magnitude of the estimated impact is relatively small. For the 29,294 patients who received

CABG surgery in states that had repealed CON in 2000, the results suggest that 29 fewer

inpatient deaths could have been avoided with CON rules. In addition, the results yield no

evidence that the volume effects associated with CON rules led to reduced mortality for patients

undergoing PTCA.

Nevertheless, repeal of CON legislation is associated with substantially higher costs per

patient. The increase in volume associated with CON regulations is predicted to reduce average

costs by 3.0 percent for PTCA patients and 6.7 percent for CABG patients. These results might

lead one to favor CON legislation, because it appears to reduce costs and either saves some lives,

or leaves mortality rates unchanged.

However, the presence of minimum volume standards may lead hospitals in CON states

to increase the number of procedures performed relative to states without CON. The predicted

increases in the total number of procedures performed (41 percent for PTCA and 18 percent for

CABG in the year 2000) are large enough to offset any potential savings resulting from lower

average costs per patient treated as a result of CON regulation. These results are consistent with

past research which has found CON regulations do not restrain expenditure growth.


27

Nevertheless, the estimates of statewide growth in PTCA procedures resulting from CON

seem extraordinarily large. I am in the process of surveying states to obtain data on minimum

volume thresholds for open heart surgery and cardiac catheterization. Preliminary investigation

indicates that minimum volume thresholds vary across states, but little change in thresholds

occurs across years within states. In future research, I aim to combine this information with data

from more states to revisit the issue of CON and statewide procedure volume.

This study does not attempt to measure the potential impact of CON rules on the

substitution of PTCA for CABG procedures, which has increased in the U.S. over time (D. M.

Cutler, R. S. Huckman 2003). However, the overwhelming majority of states with cardiac CON

rules restrict entry for both CABG and PTCA. Therefore, rates of substitution between these two

procedures smay not vary by CON status. This issue is an interesting question for future

research.

In the meantime, policy makers must carefully weigh the costs and benefits of CON

regulation. Policy makers implemented CON with the dual goals of restraining cost growth and

achieving high quality care. The centralization of care which is associated with CON may lead to

slightly lower mortality rates for CABG and lower unit costs due to economies of scale.

However, these regulations may also inadvertently increase the total number of procedures

performed. Future studies with more detailed clinical data should attempt to measure the

marginal benefit of increased rates of cardiac surgery in states which have maintained CON

legislation. The potential impact of differences in rates of surgery resulting from CON on long

term mortality has important consequences for evaluating the welfare benefits of these

regulations.


28

REFERENCES

1. Conover, C. J., Sloan, F. A. "Does removing Certificate-of-Need regulations lead to a surge in health care spending?" Journal of Health Politics, Policy and Law, 1998, 23 (3), 455-481.

2. Cutler, D. M., Huckman, R. S. "Technological development and medical productivity: the diffusion of angioplasty in New York state." Journal of Health Economics, 2003, 22 (2), 187-217.

3. Dafny, L. "Entry deterrence in hospital procedure markets: A simple model of learning-by-doing." Journal of Economics and Management Strategy, 2005, 14 (3), 513-542.

4. Dranove, D. "Measuring costs," Sloan, F. A., Valuing health care. New York, NY: Cambridge University Press, 1995, 61-75.

5. Grabowski, D. C, Ohsfeldt, R. L., Morrisey, M. A. "The effects of CON repeal on Medicaid nursing home and long-term expenditures." Inquiry, 2003, 40 (2), 146-157.

6. Hannan, E. L., Racz, M., Ryan T.J., McCallister, B. D., Johnson, L. W., Arani, D. T., Guerci, A. D., Sosa, J., Topol, E. J. "Coronary angioplasty volume-outcome relationships for hospitals and cardiologists." The Journal of the American Medical Association, March 1997, 277 (11), 892-898.

7. Hensley, S. The patient or the wallet: N.H. hospitals battle over open-heart surgery CONs. Modern Healthcare , 24. 8-25-1997.

8. Ho V. "Learning and the evolution of medical technologies: The diffusion of coronary angioplasty." Journal of Health Economics, 2002, 21 (5), 873-885.

9. Ho, V. "Evolution of the volume-outcome relationship for hospitals performing coronary angioplasty." Circulation, April 2000, 101 1806-1811.

10. Ho, V. "Letter to the Editor: Certificate of Need regulations and hospital mortality." The Journal of the American Medical Association, February 2003, 289 (5), 551.

11. Kessler, D. P., McClellan, M. B. "Is hospital competition socially wasteful?" The Quarterly Journal of Economics, 2000, 115 (2), 577-615.

12. Kimmel, S. E., Berlin, J. A., Laskey, W. K. "The relationship between coronary angioplasty procedure volume and major complications." Journal of the American Medical Association, 1995, 274 (14), 1137-1142.

13. Lanning, J. A., Morrisey, M. A., Oshfedlt, R. L. "Endogenous hospital regulation and its effects on hospital and non-hospital expenditures." Journal of Regulatory Economics, 1991, 3 137-154.


29

14. Maryland Health Care Commission. An analysis and evaluation of Certificate of Need regulation in Maryland, Working Paper: Cardiac Surgery and Therapeutic Catheterization Services. 8-18-2000. Baltimore, MD.

15. Meltzer, D., Chung, J., Basu, A. "Does competition under Medicare prospective payment selectively reduce expenditures on high-cost patients?" RAND Journal of Economics, 2002, 33 (3), 447-468.

16. Noether, M. "Competition among hospitals." Journal of Health Economics, 1988, 7 259-284.

17. Pallarito, K. Change of heart: N.J. expansion of cardiac CONs proving controversial. Modern Healthcare , 44. 7-20-1998.

18. Robinson, J. L., Nash, D. B., Moxey, E., O'Connor, J. P. "Certificate of need and the quality of cardiac surgery." American Journal of Medical Quality, 2001, 16 (5), 155-160.

19. Romano, P. S., Roos, L. L., Jollis, J. G. "Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: Differing perspectives." Journal of Clinical Epidemiology, 1993, 46 (10), 1075-1079.

20. Vaughan-Sarrazin, M. S., Hannan, E. L., Cormley, C. J., Rosenthal, G. E. "Mortality in Medicare beneficiaries following coronary artery bypass graft surgery in states with and without Certificate of Need regulation." The Journal of the American Medical Association, October 2002, 288 (15), 1859-1866.

21. Wooldridge, J. M. Introductory Econometrics: A Modern Approach. Mason, OH: South-Western College Publishing, 2000.

22. Wooldridge, J. M. Econometric analysis of cross section and panel data. Cambridge, MA: The MIT Press, 2002.

23. Wysocki, B. Care, costs and competition - Medical, business interests spar over state hospital-growth rules. The Wall Street Journal , A.4. 5-7-2002. New York.


30

Figure 1

Sample: All states in NIS dataset, except PA. CON: CT, FL, GA, HI, IA, IL, KY, MA, MD, ME, MO, NC, NJ, NY, SC, TN, VA, WA no CON: AZ, CA, CO, KS, OR, UT, TX, WI


31

Figure 2


32

Figure 3

11000

12000

13000

14000

15000

Co

st

pe

r P

atie

nt

in 2

00

0 D

olla

rs

1988 1990 1992 1994 1996 1998 2000

Year

no CON CON

Average Cost per PTCA by CON status

20000

25000

30000

35000

Co

st

pe

r P

atie

nt

in 2

00

0 D

olla

rs

1988 1990 1992 1994 1996 1998 2000

Year

no CON CON

Average Cost per CABG by CON status

Sample: All states in NIS dataset, except PA. CON: CT, FL, GA, HI, IA, IL, KY, MA, MD, ME, MO, NC, NJ, NY, SC, TN, VA, WA no CON: AZ, CA, CO, KS, OR, UT, TX, WI


33

Figure 4

10000

10500

11000

11500

Cost

per

Patient

in 2

000 D

olla

rs

1988 1990 1992 1994 1996 1998 2000

Year

Average Cost per PTCA Patient for Pennsylvania

(PA removed CON in 1996)

20000

25000

30000

35000

Cost

per

Patient

in 2

000 D

olla

rs

1988 1990 1992 1994 1996 1998 2000

Year

Average Cost per CABG Patient for Pennsylvania

(PA removed CON in 1996)


34

Table 1: Regression Estimates, Determinants of Hospital Procedure Volumea

_________ln(PTCA Volume)__________ ________ln(CABG Volume)__________ coefficient t-statistic coefficient t-statistic

yrsnoCON -0.033** (-2.43) -0.022* (-1.81) log(pop65+/sqmile) 1.166** (2.09) 1.267*** (2.48) HMO penetration rate -0.036 (-0.22) -0.109 (-0.87) state smoking rate 1.176 (0.94) -0.507 (-0.55) 1989 0.075 (1.39) 0.017 (0.29) 1990 0.297*** (4.15) 0.096 (1.44) 1991 0.407*** (5.13) 0.139** (1.95) 1992 0.540*** (6.21) 0.163** (2.00) 1993 0.626*** (6.63) 0.189** (2.19) 1994 0.727*** (7.60) 0.253*** (2.79) 1995 0.849*** (7.89) 0.302*** (3.06) 1996 0.882*** (7.85) 0.385*** (3.67) 1997 0.993*** (8.52) 0.390*** (3.57) 1998 1.092*** (9.05) 0.381*** (3.32) 1999 1.155*** (9.64) 0.332*** (2.85) 2000 1.359*** (10.20) 0.381*** (3.05) constant -0.230 (-0.10) 0.102 (0.05)

Obs. 2367 2195 # hospitals 561 497 _________________________________________________________________________________________________________

aRegressions include hospital-specific fixed effects. All standard errors are estimated with robust standard errors, which account for clustering of patients within hospitals.

*statistical significance at the 10% level. **statistical significance at the 5% level. ***statistical significance at the 1% level.


35

Table 2: Regression Estimates, Determinants of Risk-Adjusted Mortality Rates PTCA CABG OLS Random Effects Fixed Effects OLS Random Effects Fixed Effects

log(Volume)ht -0.0037*** (-2.54) -0.004** (-2.38) -0.0004 (-0.22) -0.006*** (-3.34) -0.004*** (-2.62) -0.004 (-1.27) 1989 -0.0001 (-0.07) 0.0004 (0.26) 0.0004 (0.25) -0.0009 (-0.35) -0.001 (-0.57) -0.001 (-0.53) 1990 0.00002 (0.01) 0.001 (0.41) 0.0001 (0.04) -0.007** (-2.27) -0.006** (-2.43) -0.006** (-1.99) 1991 0.002 (0.95) 0.002 (1.22) 0.001 (0.52) -0.009*** (-2.97) -0.008** (-3.13) -0.008*** (-2.48) 1992 0.005 (1.41) 0.005 (1.62) 0.002 (0.98) -0.010*** (-3.20) -0.010*** (-3.28) -0.009** (-2.45) 1993 0.003* (1.73) 0.004** (2.23) 0.003 (1.23) -0.013*** (-4.30) -0.011*** (-4.33) -0.010*** (-3.16) 1994 0.005** (2.35) 0.006*** (2.93) 0.005* (1.87) -0.012*** (-3.78) -0.011*** (-4.09) -0.010*** (-2.94) 1995 0.007*** (3.25) 0.008*** (3.63) 0.005** (2.08) -0.013*** (-3.97) -0.012*** (-4.51) -0.011*** (-3.11) 1996 0.006*** (2.77) 0.007*** (3.00) 0.004 (1.35) -0.014*** (-4.56) -0.013*** (-5.11) -0.012*** (-3.52) 1997 0.005*** (2.53) 0.005*** (2.83) 0.002 (1.01) -0.015*** (-4.71) -0.014*** (-5.01) -0.012*** (-3.41) 1998 0.004* (1.92) 0.004** (2.08) -0.001 (-0.20) -0.017*** (-4.80) -0.015*** (-4.63) -0.013*** (-2.90) 1999 0.007* (1.66) 0.007* (1.75) 0.003 (0.78) -0.015*** (-4.56) -0.013*** (-4.15) -0.010** (-2.16) 2000 0.002 (1.13) 0.003 (1.33) -0.0004 (-0.15) -0.018*** (-5.45) -0.015*** (-5.35) -0.012*** (-3.11) constant 0.035*** (4.66) 0.035*** (4.30) 0.018** (2.10) 0.089*** (8.10) 0.078*** (8.19) 0.079*** (4.45)

Obs. 2367

2195

# hospitals 561 497

*statistical significance at 10% level.

**statistical significance at 5% level. ***statistical significance at 1% level.


36

Table 3: Regression Estimates of Determinants of Log(Average Costs per Procedure) PTCA CABG Coefficient t-statistic Coefficient t-statistic Log(volume) -0.070*** (-2.78)

Volume -0.001*** (-2.85) Volume 2 5.54e-07** (2.43) Volume 3 -1.44e-10** (-2.11) 1989 0.018 (1.04) 0.030* (1.76) 1990 0.021 (1.01) 0.016 (0.94) 1991 0.061*** (2.64) 0.055*** (3.11) 1992 0.019 (0.76) 0.011 (0.59) 1993 0.008 (0.28) -0.015 (-0.69) 1994 -0.012 (-0.39) -0.063*** (-2.65) 1995 -0.009 (-0.28) -0.137*** (-5.02) 1996 -0.105*** (-3.28) -0.192*** (-7.91) 1997 -0.166*** (-5.02) -0.227*** (-9.34) 1998 -0.156*** (-4.44) -0.218*** (-8.41) 1999 -0.178*** (-4.70) -0.228*** (-9.07) 2000 -0.185*** (-4.34) -0.222*** (-7.79) Age 65-69 y 0.044*** (22.51) 0.060*** (29.07) Age 70-74 y 0.066*** (23.41) 0.108*** (41.42) Age 75-79 y 0.098*** (28.72) 0.156*** (47.24) Age 80-84 y 0.132*** (25.59) 0.211*** (41.38) Age 85 y 0.173*** (20.50) 0.259*** (35.02) Female 0.035*** (17.52) 0.069*** (35.29) Black 0.015** (2.30) 0.0391*** (6.89) AMI 0.282*** (48.19) 0.142*** (42.43) Emergency admit 0.242*** (26.74) 0.164*** (26.73) Transfer -0.001 (-0.06) 0.025*** (4.33) Prior AMI -0.041*** (-10.86) -0.114*** (-30.22) Peripheral Vascular disease 0.086*** (19.03) 0.016*** (4.02) Dementia 0.066*** (5.16) -0.005 (-0.31) COPD 0.108*** (26.91) 0.088*** (22.79) Rheumatologic disease 0.036*** (5.69) -0.039*** (-6.92) Liver disease (mild) 0.088*** (4.89) 0.148*** (7.98) Diabetes (mild/moderate) 0.006** (2.24) -0.046*** (-16.59) Diabetes w/complications 0.172*** (28.54) 0.064*** (13.27) Kidney Disease 0.381*** (31.33) 0.465*** (32.11) Any malignancy 0.039*** (6.37) -0.030*** (-7.30) Liver disease (moderate/severe) 0.382*** (7.93) 0.490*** (12.02)

Metastatic solid tumor 0.278*** (18.82) 0.145*** (8.44) Multivessel PTCA 0.115*** (25.75)

Stent Insertion 0.173*** (17.01)

Constant 9.43*** (63.62) 10.193*** (168.49) Number of obs. 923725 743172

*statistical significance at 10% level; **statistical significance at 5% level; ***statistical significance at 1% level


37

Table 4: Regression Estimates, Determinants of Statewide Procedure Volume ln(PTCA Volume) ln(CABG Volume) Fixed Effects 1 -Difference Fixed Effects 1 -Difference yrsnoCON -0.038*** (-3.82) -0.022 (-1.69) -0.016*** (-3.54) -0.014* (-1.78) log(pop65+/sqmile) 0.376 (0.54) 0.307 (0.48) 1.194*** (4.81) 1.482*** (2.89) HMO penetration rate

-0.307 (-0.99) 0.027 (0.15) 0.268 (1.22) 0.184 (1.31)

smoking rate 1.228 (1.47) 0.828** (2.38) 0.757 (1.35) 0.318 (0.96) per capita income 3.0e-.05 (1.40) 1.71e-05* (1.89) 1.37e-06 (0.14) -2.83e-08 (-0.01) rate uninsured -0.293 (-0.32) -0.094 (-0.43) -0.758 (-1.38) -0.132 (-0.47) 1989 0.108*** (2.43) 0.009** (0.27) 1990 0.294*** (3.86) 0.070 (1.45) 0.084** (2.42) 0.058* (1.82) 1991 0.414*** (4.61) 0.010 (0.30) 0.138*** (2.86) 0.058** (2.11) 1992 0.532*** (4.38) -0.003 (-0.09) 0.217*** (4.20) 0.058 (1.57) 1993 0.610*** (4.35) -0.042* (-1.42) 0.187*** (2.93) -0.039 (-1.15) 1994 0.721*** (4.98) -0.008 (-0.36) 0.214*** (3.64) 0.028 (0.81) 1995 0.841*** (5.10) -0.003 (-0.12) 0.287*** (4.06) 0.062** (2.29) 1996 0.918*** (4.61) -0.044 (-1.40) 0.340*** (4.42) 0.053* (1.90) 1997 0.972*** (4.20) -0.081** (-2.69) 0.329*** (3.60) -0.013 (-0.50) 1998 1.043*** (3.94) -0.043 (-1.57) 0.289*** (2.97) -0.038 (-1.19) 1999 1.058*** (3.83) -0.094*** (-2.72) 0.266*** (2.42) -0.025 (-0.83) 2000 1.090*** (3.49) 0.063** (-2.52) 0.231* (1.92) -0.028 (-0.74) 2001 1.163*** (3.55) -0.031 (-0.90) 0.174 (1.35) -0.060** (-2.13) 2002 1.210*** (3.54) -0.053** (-2.06) 0.125 (0.97) -0.053 (-1.66) constant 6.446 (3.01) 0.118*** (3.86) 4.939*** (6.23) 0.001 (0.03)

Exogeneity test F(6,20)=0.81 F(6,20)=2.75** F(6,20)=13.60*** F(6,20)=1.93

Obs. 146 125 146 125 # states 21 21 *statistical significance at 10% level; **statistical significance at 5% level; ***statistical significance at 1% level.


38

APPENDIX A Reprinted from Working Paper: Cardiac Surgery and Therapeutic Catheterization Services, Maryland Health Care Commission*

Table B1. Status of CON Regulation of Cardiovascular Services by State

CON Program Cardiac Covered (28) Cardiac Not Covered (8) No CON Program (15)

Alabama Alaska Connecticut Delaware District of Columbia Florida Georgia Hawaii Illinois Iowa Kentucky Maine Maryland Massachusetts Arizona (1985) Michigan California (1987) Mississippi Colorado (1987) Missouri Idaho (1983) New Hampshire Indiana (1998) New Jersey Kansas (1985) New York Louisiana North Carolina Arkansas Minnesota (1984) Rhode Island Montana New Mexico (1983) South Carolina Nebraska North Dakota (1995) Tennessee Nevada Pennsylvania (1996) Vermont Ohio South Dakota (1988) Virginia Oklahoma Texas (1985) Washington Oregon Utah (1984) West Virginia Wisconsin Wyoming (1985) *States in the AHRQ NIS dataset are italicized.


39

APPENDIX B Summary Statistics

Variable Obs Mean Std. Dev. Min Max

Year 2389 1994 3.66 1988 2000 Hospital PTCA Volume 2367 450.69 431.49 6 3035 Hospital CABG Volume 2195 381.18 321.64 6 2195 yrsnoCON 2389 3.10 4.68 0 17 County Pop 65+/sq mile 2187 272.93 571.35 8.96 6760.21 County HMO pen. rate 2389 0.26 0.16 0 0.91 State smoking rate 2389 0.23 0.02 0.13 0.29 State per capita income 2389 24240.20 7119.49 12027 92378.48 State percent uninsured 2389 0.17 0.05 0.08 0.28 RAMR (PTCA) 2367 0.018 0.024 0 0.73 RAMR (CABG) 2195 0.046 0.026 0 0.31 Cost per Patient (PTCA) 930,327 $11,830 $9,983 $1001 $798,945 Cost per Patient (CABG) 744,156 $27,180 $21,748 $1003 $1,696,385 Population figures were obtained from the U.S. census bureau. HMO penetration rates were obtained from Douglas Wholey at the University of Minnesota. Smoking rates were obtained from the Center for Disease Control’s Behavioral Risk Factor Surveillance System. State per capita income was obtained from the Bureau of Economic Analysis. The percent uninsured was obtained from the March Current Population Surveys.


40

Appendix C Alternative Specifications of the Association between Years since CON Repeal and Procedure Volume

Hospital Procedure Volume Statewide Procedure Volume ln(PTCA Volume) ln(CABG Volume) ln(PTCA Volume) ln(CABG Volume)

coefficient t-statistic coefficient t-statistic

coefficient t-statistic coefficient t-statistic

yrsnoCON -0.023 (-0.65) -0.012 (-0.48) -0.063*** (-3.06) -0.020* (-1.84)

yrsnoCON2 -0.001 (-0.27) -0.001 (-0.41) 0.001 (1.36) -0.0003 (0.61) yrsnoCON 0.044 (0.65) -0.037 (-0.79) -0.037 (-1.34) -0.003 (-0.22) yrsnoCON2 -0.011 (-1.06) 0.003 (0.47) -0.002 (-0.65) -0.002 (-1.22) yrsnoCON3 0.0004 (0.96) -0.0002 (-0.54) 0.0001 (1.16) 0.0001 (1.33) tertile 1‡ -0.021 (-0.87) -0.028 (-1.58) -0.046*** (-3.98) -0.020** (-2.75) tertile 2 -0.022 (-0.65) 0.019 (0.77) 0.006 (0.31) 0.004 (0.40) tertile 3 0.024 (0.42) -0.039 (-1.10) 0.042 (1.47) 0.010 (0.58) F-test† F= 0.22 p= 0.806 F= 0.65 p= 0.524 F= 1.51 p= 0.245 F= 0.35 p= 0.711 quartile 1± -0.014 (-0.47) -0.017 (-0.82) -0.045*** (-3.33) -0.018** (-2.33) quartile 2 -0.023 (-0.43) -0.023 (-0.71) -0.003 (-0.12) -0.004 (-0.34) quartile 3 -0.004 (-0.08) 0.043 (1.42) 0.032** (2.81) 0.017 (1.31) quartile 4 0.022 (0.25) -0.077 (-1.53) 0.022 (0.86) -0.002 (-0.11) F-test† F= 0.17 p= 0.915 F= 1.02 p= 0.382 F= 4.55 p= 0.014 F= 0.66 p= 0.588 *statistical significance at 10% level; **statistical significance at 5% level; ***statistical significance at 1% level. §All alternative specifications of yrsnoCON include the same set of explanatory variables reported in Tables 1 and 4. ‡ Tertile cutoffs are 8 and 14 years without CON for hospital procedure volume and 6 and 11 years without CON for statewide procedure volume. ±Quartile cutoffs are 7, 12, and 15 years without CON for hospital procedure volume and 5, 8, and 12 years without CON for statewide procedure volume. †F-test for the joint significance of the coefficients on the 2 highest tertiles or 3 highest quartiles.

oes ertificate of eed ffect ardiac utcomes and osts v h · 2013. 8. 6. · the james a. baker iii...

Documents